基于电力大数据的用户行为分析及可视化技术应用

发布时间：2018-03-18 06:05

本文选题：智能电网　切入点：大数据　出处：《华北电力大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着电网规模的不断扩大,加之智能终端设备的安装覆盖率逐年提高,使得电网的业务数据正以爆炸性的趋势增长,如何利用数据挖掘技术,在海量的电网数据中挖掘出有价值的信息,成为当前电力系统分析中的一个挑战性问题。另一方面,随着市场经济的发展,电网公司正在逐渐转化为经营型企业,如何对用电用户的行为进行分析和预测,并为其提供个性化的电力服务,成为电网公司亟待解决的问题。因此,对电力大数据进行分析和研究,将有利于提高电网公司的经营管理水平和建立坚强的智能电网。本文研究了国外大数据的发展概况以及我国大数据的发展现状和行业动态,重点研究了当今电力大数据现状以及天津市电力公司对业务场景的需求分析,提出了基于电力大数据的用户行为分析及可视化展示平台,该平台实现了用电客户分类和用户用电量预测功能,并对分析结果进行可视化展示。本文首先通过分析目标用电客户的用电行为、用电习惯和用电规律,将数据挖掘与用电客户分类相结合,完成了电力客户分类整体模型的设计和指标体系的构建。将聚类算法与Hadoop分布式处理框架相结合,给出了基于Hadoop的并行化K-means算法,利用电网公司营销侧数据,通过分析电力客户的用电量情况、信用情况和价值创造等,实现了用电客户分类的功能。用户的用电量行为在具有自身规律的同时,也受着外界因素的影响。将关联规则挖掘算法与Hadoop分布式处理框架相结合,给出了基于Hadoop的并行化Apriori算法,结合会对用户用电量行为带来影响的经济、气温等影响因素,通过分析相关因素对用户用电量的影响和关联性,实现了用户用电量预测功能。最后通过ExtJS技术,将挖掘出的用户行为进行可视化展示,辅助电网公司提高自身的业务水平和经营管理水平,使电网公司能够更有针对性的开展市场营销业务并提高其客户服务能力。
[Abstract]:With the continuous expansion of power grid scale and the increasing installation coverage rate of smart terminal equipment, the service data of power grid is increasing explosively. How to use data mining technology, Mining valuable information from massive power grid data has become a challenging problem in power system analysis. On the other hand, with the development of market economy, power grid companies are gradually transforming into operating enterprises. How to analyze and predict the behavior of power users and provide them with personalized power services has become an urgent problem for power grid companies. It will help to improve the management level of power grid companies and establish a strong smart grid. This paper studies the general situation of big data's development abroad, the current situation of the development and the industry trends of our country big data. This paper mainly studies the current situation of power big data and the demand analysis of Tianjin Electric Power Company to the business scene, and puts forward the user behavior analysis and visual display platform based on power big data. The platform realizes the function of customer classification and power consumption prediction, and visualizes the analysis results. Firstly, this paper analyzes the behavior, habits and rules of electricity consumption of the target customers. Combining data mining with customer classification, the whole model of power customer classification is designed and the index system is constructed. By combining clustering algorithm with Hadoop distributed processing framework, a parallel K-means algorithm based on Hadoop is presented. By analyzing the electricity consumption, credit and value creation of electric power customers, the function of customer classification is realized by using the marketing side data of power grid company. At the same time, the electricity consumption behavior of users has its own law. By combining association rule mining algorithm with Hadoop distributed processing framework, a parallel Apriori algorithm based on Hadoop is presented. By analyzing the influence and correlation of the related factors on the user's electricity consumption, the function of user's electricity consumption prediction is realized. Finally, the user's behavior is visualized through ExtJS technology. Assist the grid company to improve its own business level and management level, so that the grid company can carry out marketing business and improve its customer service ability.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TM73

【参考文献】